Figure 1: Democratic Vote Share and Unemployment Growth Note: Counties with at least 50,000 residents in 2008 according to the Census bureau are included in the figure. To control for demographics, vote is residualized using percent white, percent black, percent Hispanic, and income, income squared, and income cubed (see the SI for details). That is, it presents "component plus residual" plots that residualize the dependent variable for variation explained by the percent white, black, and Hispanic as well as for income, income squared, and income cubed. For details and the full models, see the supporting information. The lpoly curve is weighted by votes cast in the county. Two outliers for unemployment are dropped from the top plot. 2006 Percent 15 10 5 Percent 0 15 2007 10 5 0 2008 Percent 10 5 Percent 0 10 2009 5 Percent 0 10 2010 5 0 0 10 20 30 Mortgages 90 days delinquent % Figure 2: Histograms of the Mortgage Delinquency Rates for California Zip Codes Note: Plot shows zip codes with approximately 250 registered voters or more. Figure 3: Democratic Vote Share and Change in Mortgage Delinquency Note: To control for demographics, vote is residualized using a regression of the change in Democratic vote share on a set of demographic variables (see Table 1, Column 3). Stata's lpoly command produced the best-fit curve and is weighted by the number of zip-code registered voters. Table 1: ZIP Code Mortgage Delinquencies and Presidential Voting, 2008 Dependent variable: Democratic share of two-party presidential vote, 2008 Level of analysis: Zip code Change in share of mortgages 90+ days delinquent, 2006-2008 Democratic vote share, 2004 (1) (2) (3) 7.68*** (0.87) 0.90*** (0.01) 6.73*** (1.01) 0.90*** (0.011) 0.011 (0.018) 0.003 (0.014) 0.011 (0.011) 9.87*** (0.82) 9.86*** (1.20) 7.68*** (1.08) 0.89*** (0.010) 0.026 (0.018) -0.0017 (0.013) 0.030** (0.014) 0.062*** (0.012) -0.00024*** (0.000062) 2.3e-07*** (6.8e-08) 7.12*** (1.42) Percent black, 2000 Percent white, 2000 Percent hispanic, 2000 Average income, 2001 Average income squared, 2001 Average income cubed, 2001 Constant 1,386 1,386 1,386 Observations 0.983 0.983 0.985 R-squared Note: Since we control for Democratic vote share in 2004 (the lagged dependent variable), this is a model of change in Democratic vote share (see Finkel 1995). Robust standard errors clustered at the county level. Weighted by the number of zip-code registered voters. *** p<0.01, ** p<0.05, * p<0.1 Table 2: Robustness of Local Economy's Effect on Presidential Elections for Delinquencies Coefficient estimate (SE) Baseline Specification Change in share of mortgages 90+ days delinquent, 2006-2008 (Table 1, col. 3) 7.7 (1.1) A. Different Delinquency Measures in Zip Code Change in share of mortgages 90+ days delinquent, 2007-2008 15.3 (2.3) Change in share of mortgages 30+ days delinquent, 2006-2008 8.1 (1.2) Change in share of mortgages 30+ days delinquent, 2007-2008 18.0 (3.1) B. Controlling for Migration Controlling for black, hispanic, and white migration 5.4 (1.1) C. Breaking the Effect Down by Previous Vote in the Zip Code 1st quartile of 2004 Democratic vote 10.1 (2.1) 2nd quartile of 2004 Democratic vote 10.0 (1.6) 3rd quartile of 2004 Democratic vote 5.0 (1.2) 4th quartile of 2004 Democratic vote -0.5 (1.1) D. Breaking the Effect Down by High and Low Population Zip Codes Below the median zip-code population (no weights) 3.1 (1.4) Above the median zip-code population (no weights) 8.1 (1.4) E. Placebo Tests Using Post-Election Delinquencies in Zip Code Change in share of mortgages 90+ days delinquent, 2008-2009 1.0 (2.8) Note: All estimates are based on Table 1, column 3 model except where noted. Robust standard errors clustered at the county level, except where noted. Weighted by the total number of voters, except where noted. Change in mortgage delinquencies is rescaled to vary between zero and one. All estimates are statistically significant at p<0.01 except for the "below the median zip-code population" and "placebo test" estimates. Table 3: The Local Economy and Presidential Vote, 1992-2012 Dependent variable: Democratic share of two-party presidential vote Level of analysis: County Expected sign: QCEW wage and employment growth*Democratic incumbent (1) Population measure (2) Population measure (3) Population measure (4) Sample measure (5) Sample measure + + + - + 9.82*** 6.42*** 5.95*** (1.26) (0.99) (0.97) BLS Unemployment growth*Democratic incumbent 1.09 (1.49) SAIPE Income growth*Democratic incumbent 0.86 (1.12) Controls Previous election Democratic vote share 1.03*** (0.0051) 1988 per-capita income 0.91*** 0.91*** 0.92*** (0.0071) (0.0072) (0.0083) 6.29 7.34 6.30 (5.34) (5.47) (6.93) -3.05* -3.46* -3.41 (1.76) (1.81) (2.29) 0.10 0.40** 0.44** 0.46** (0.15) (0.17) (0.18) (0.23) 5.61*** 5.47*** 8.15*** (1.23) (1.43) 4.40*** (1.22) (1.27) 5.99*** (1.25) 4.88*** 4.76*** 4.60*** (1.00) (1.02) 1.00*** (0.0057) -3.25 (4.84) 1988 per-capita income squared 0.014 (1.53) 1988 per-capita income cubed Percent black Percent white -5.78*** Percent hispanic (1.34) 11.0*** 15.3*** 15.0*** 11.1** (1.08) -0.27 (0.69) (4.46) (4.48) -4.71 -5.68 Year effects? Y Y Y Y Y State fixed effects? N Y Y Y Y Observations 18,511 18,510 18,510 18,510 12,357 R-squared 0.932 0.94 0.946 0.946 0.959 Constant Note: The first three columns of this table show that the population-based QCEW measure of the local economy strongly predicts incumbent president vote share, while columns 4 and 5 showed that sample-based measures do not. We rescale the local economy measures to vary between zero and one. Since we control for Democratic vote share (the lagged dependent variable), this is a model of change in Democratic vote share (see Finkel 1995). We do not include the main effect of Democratic incumbent because its effect is captured by the year fixed effects. Robust standard errors clustered at the county level. Weighted by the total number of voters. *** p<0.01, ** p<0.05, * p<0.1 Table 4: Robustness of Local Economy's Effect on Presidential Elections QCEW wage and employment growth Estimate (SE) BLS Unemployment growth SAIPE Income growth Population measure Sample measure Sample measure + - + From Table 3, Columns 3, 4, and 5 6.0 (1.0)*** 1.1(1.5) 0.9(1.1) All estimated in same sample (2000-2012) 9.2 (1.1)*** 0.7(1.4) 0.9(1.1) A. Testing for Sensitivity to Individual Elections Largest effect excluding elections individually and reestimating 7.3(1.1)*** -3.9(2.0)* 2.1(1.4)* Smallest effect excluding elections individually and reestimating 4.1(0.8)*** -0.8(1.6) -0.3(1.1) All controls interacted with election-year effects 5.6(0.9)*** 0.7(1.1) 1.0(1.0) Controlling for lagged annual wage 7.1(1.1)*** -0.7(1.5) 0.8(1.1) Controlling for lagged annual wage, squared, and cubed 6.7(1.1)*** -0.4(1.6) 0.6(1.1) Controlling for population growth 5.7(1.0)*** 1.2(1.5) 0.8(1.1) Counties with 25,000 voters or more (no weights) 5.0(1.1)*** -0.5(1.5) -0.5(1.1) Counties with 50,000 voters or more (no weights) 5.6(1.3)*** 0.4(1.9) -0.7(1.4) 2.4(0.9)*** 2.4(1.6) 1.6(1.2)* Expected sign: Baseline Specification B. Controlling for Additional Variables C. Dropping Smaller Counties and Not Using Weights D. Varying the Economic Measure Not mean deviated Employment growth 0.8(1.0) Wage growth 4.7(0.9)*** Average of wage and empl. growth for year before election 5.9(1.0)*** E. Placebo Tests Using Post-Election Economic Conditions Regression from Table 3, Column 2 0.8(0.6) -0.0(0.1) 0.1(0.0) Regression from Table 3, Column 3 0.5(0.6) -0.1(0.1) 0.1(0.0) Note: All models are based on Table 3, columns 3, 4, and 5 except where noted. Growth measures are rescaled to vary between zero and one. Robust standard errors clustered at the county level. Weighted by the total number of voters in the county, except where specified. *** p<0.01, ** p<0.05, * p<0.1 only noted for expected direction (signs).